Second test set

98 - Flow controller: the minimum transaction time for the control through Modbus Tester interface equals 30 ms from Groningen and rises to about 100 ms from Turin. The minimum sample times for the measurements through the implemented code in Matlab are 50 ms from Groningen and 140 ms from Turin²⁶. This slight delay is due to the fact that the instrument communication is based on a normal Modbus TCP/IP protocol [14,24], so that the query from the remote PC must be received before the instrument could send back the reply.

In conclusion, the latency introduced by the VPN communication only slightly affects the remote measurements in terms of regularity of the time step for the power analyzer and minimum sample time for the flow controller. Also the transaction time for the remote control of the flow controller setpoint is marginally influenced. Nevertheless, the delays introduced in the entire control chain are minimal and do not prevent the use of the multi-site setup for research purposes.

99

Fig. 76 – Simulink model subsystem: Flow controller Modbus writing

Fig. 77 – Simulink model subsystem: Flow controller Modbus reading

Fig. 78 – Simulink model subsystem: Power analyzer TCP reading

With this model, two identical tests were performed by giving the flow controller the same setpoint already described in the previous section, first locally from Groningen and then remotely from Turin.

100 The results obtained for the three-phase active power are shown in Figure 79 (model run from Groningen), and in Figure 80 (model run from Turin). All the considerations on the electrolyser dynamic behavior presented in the above section are still valid here. From both graphs, by focusing of phase 1, it can be observed: the electrolyser start-up phase comprising generate-vent and pressurize-storage states; then the steady-states at 25, 50, 75 and 100% of full-load, with a very fast dynamic from one steady-state to the next one. The only main difference highlighted in these cases is that the ranges of power variation are consistently reduced if compared with the results of the previous experimental set. The measured values at each time step appear to be only an average of the real data within the considered interval.

Fig. 79 – Active power withdrawn by the electrolyser - Local measurements (from Simulink)

Fig. 80 – Active power withdrawn by the electrolyser - RHIL measurements (from Simulink)

This is due to the fact that, within the Simulink environment, the instruments cannot be read at the minimum sample time available as in the previous cases, but this has to be necessarily higher because of the further latency introduced by the simulation. In fact, it is worth to notice

101 that the Simulink block Matlab Function works by interacting with Matlab while running the simulation. At each simulation time step the model is momentarily paused, the Matlab Function blocks interface with the respective Matlab file and this runs the code to read/write data from and to the instruments. Finally, when the data replies are received, they are transferred from Matlab to Simulink and the simulation starts again moving to the subsequent time step.

This interaction increased the minimum simulation time step to 400 ms when the model was run from Groningen. When the same model was run from Turin, the latency due to the communication Matlab-Simulink in addition to the delay due to the VPN communication further raised the minimum simulation time step to 600 ms²⁷. In both cases, this is the minimum fixed-step time it was possible to set. In fact, since the simulation time step also identify the time period in which new values are requested and received from the instruments, a lower value would create too much data traffic which would slow down the communication and eventually lead to failure in the instruments and simulation crash.

At last, the measurements of H2 output flow are presented in Figure 81 (model run from Groningen) and in Figure 82 (model run from Turin). The two graphs are barely distinguishable from the respective graphs obtained during the previous test case. In fact, unlike the power read from the power analyzer, the H2 flow measured by the flow controller is generally not influenced by the huge increment in time step due to the simulation latency. This is because the flow is constant during the steady-states and does not present the variation typically analyzed in the power behavior. Only the accuracy in capturing the flow dynamic, lasting less than 1 s, might be influenced by the higher time step.

Fig. 81 – Production of H2 - Local measurements (from Simulink)

27 The simulation was performed as a normal non-real-time simulation, with the simulation time step set to 25 ms. The simulation duration was timed and the time step scaled up to real-time by multiplying for the conversion factor.

102

Fig. 82 – Production of H2 - RHIL measurements (from Simulink)

In conclusion, it is possible to manage the measurement and control chain of the electrolyser from within a simulation running locally and – what is more important, remotely, but this leads to limits in terms of minimum sample time for reading the instruments and minimum simulation time step. In particular, this influences the accuracy in reading the power from the power analyzer, whereas the reading of the flow controller is only marginally affected with regard to the capture of the flow dynamics. Improvements in the method of integrating the measuring and control devices into Simulink may lead to a lower simulation time step (for example by developing appropriate Simulink S-function blocks for interacting with the instruments without any need to interface with the Matlab environment).

103

Conclusion and future work

This thesis work explained in detail all the technical aspects concerning the creation of a multi-site laboratory setup for measuring and control of energy transition sources for PtG investigations, both from local and remote locations. The activities involved the installation and test of the different infrastructure parts, as well as their connections. In particular, the main systems developed so far count in:

 Local electrical grid, which can be managed either connected to the main grid or in island. This infrastructure aims to replicate a micro-grid environment and allows the connection of different components, considering both loads and generation.

 Energy transition sources, comprising a PV-field (energy source) of total nominal power equal to 4.16 kW and a PEM electrolyser (load – energy storage) of maximum power equal to 8 kW.

 Measurement infrastructure, mainly composed of power analyser, able to properly measure with high sampling rate electrical variables (i.e., active and reactive power, current, voltage and so on), to capture the dynamic behavior of the connected resource. Furthermore, proper sensors are available to monitor the hydrogen production process parameters, whereas temperature sensors and a pyranometer exist for monitoring weather conditions surrounding the PV-field.

 Control system, up to now consisting only in the flow controller, that allows to modify the hydrogen production of the electrolyser which in turn affects the unit power consumption. In the future, the control will be possible also by directly acting on the system electrical inputs (AC side) through a power amplifier, which is able to replicate the network conditions given manually as setpoints or simulated in another platform (i.e., Real-Time Simulator).

 Communication infrastructure, based on high-speed Ethernet and Real-Time network, interconnecting the various measuring and regulation devices for receiving the collected measurements and sending the control signals to change the operating conditions of the electrolyser in quasi real-time.

 VPN connection, joining the two distant laboratories in a multi-site laboratory layout, with the purpose of sharing existing facilities and expertise for interdisciplinary studies carried out by means of remote tests on local resources.

The test field developed was applied to build experimental-based models of PV panels and electrolyser. Appropriate data were collected from the resources installed on site through the related measurement equipment. These data were further processed by maintaining the meaning of the overall energy and finally used to implement Simulink models suitable for integration into big scale power grid simulations.

The overall communication infrastructure was tested to demonstrate the possibility to send control signals and collect measurements over the VPN connection. The control and measuring devices of the electrolyser were integrated into a simulation and the model obtained was run both locally and remotely by giving the same setpoints. Despite slight delays

104 introduced by the connection latency, the electrolyser dynamic behavior was captured in the same way from both the approaches. This opens the possibility to use the entire communication chain and the collected data for research purposes.

The next steps will be focused on the implementation of the Simulink model of a large-scale power grid, running on the Real-Time Simulator. This will be eventually decoupled into subsystems to run on different Real-Time Targets in order to realize a “geographically distributed real-time co-simulation”. The setpoints will be used to control the electrical variables of the electrolyser trough the power amplifier, already installed in the lab setup and for which a model is under design. Proper tests will be carried out by employing the entire multi-site infrastructure in Remote Power Hardware-In-the-Loop.

105

Appendix A

The aim of this appendix is providing some helpful instructions on how to use the PPA1500 power analyzer and DataLogger software that comes with it. In particular, individual features and options of this software which have been useful during the performed experiments are illustrated.

First of all, it is worth to say that DataLogger is able to connect to the power analyzer via RS232, USB and LAN. The program comprises all measurement modes which reflect instrument operating options. It also allows to export text files in CSV format or directly to Microsoft Excel.